| Summary: | sci-libs/tensorflow-1.9.0_rc1-r1 fails to compile. I cannot specify my CUDA SDK version. | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | ZongyuZ <zzy2529420793> |
| Component: | Current packages | Assignee: | Jason Zaman <perfinion> |
| Status: | RESOLVED DUPLICATE | ||
| Severity: | normal | ||
| Priority: | Normal | ||
| Version: | unspecified | ||
| Hardware: | AMD64 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: |
emerge --info output
build.log |
||
Created attachment 537508 [details]
build.log
Can you try _rc1-r2? I fixed cuda in that one. If you need to specify the exact cuda version you can set this in your make.conf TF_CUDA_COMPUTE_CAPABILITIES="2.0" (the compute capability number comes from: https://developer.nvidia.com/cuda-gpus) *** This bug has been marked as a duplicate of bug 656582 *** |
Created attachment 537506 [details] emerge --info output I tried to emerge tensorflow with a simple "emerge --ask tensorflow", but it failed. It seems like the problem is that I can't specify my own cuda-sdk version, which is 8.0.61, and the default version while compiling tensorflow is cuda-9.0. By the way, my GPU, Tesla C2050, doesn't support cuda-9.0, so I can't simply update my cuda-sdk. But I still want to install tensorflow with portage, so what can I do? Thanks!